Fault diagnosis of roller bearing using parameter evaluation technique and multi-class support vector machine

Roller bearing is one of the vital parts of a rotating machine. Bearing failure can result in serious damage of the machine. This paper aims to develop a bearing fault diagnosis method using parameter evaluation technique to improve the diagnosis accuracy. The parameter evaluation technique is used to select five features that are used as predictors in multi-class support vector machine (SVM) classification. The purpose of this feature reduction was to avoid the curse of dimensionality and to increase the accuracy of the diagnosis. The diagnosis process was performed by classification of bearing states using one-against-one method multi-class SVM. Three types of kernel functions i.e., linear, polynomial, and Gaussian RBF were used in the SVM classification. The bearing conditions which is diagnosed in this paper were normal bearing, inner race fault, and outer race fault conditions. As a result, the classification performance of multiclass SVM using five selected features as the parameter have excellent performance in predict the bearing conditions data for all types of kernel functions.